Project Summary Nearly 30% of the two million Americans suffering from epilepsy continue to have seizures despite treatment. There is now a growing acceptance of stimulation devices as a mean for therapeutic neuromodulation. To provide sophisticated feedback stimulation, one can either respond early into the seizure in order to minimize its impact and spread, or one can respond before the seizure to some other state. Identification of a suitable preseizure state has been a central theme for quite a while that now falls under the rubric of seizure prediction. Identification of such a preseizure state could both indicate when to stimulate in order to avert an oncoming seizure. But, the experience in the seizure prediction community has been that all measures used so far yield a significant false detection rate for significant levels of sensitivity. This work will be performed in the tetanus toxin model for temporal lobe epilepsy. We have developed a system for applying low frequency electrical stimulation for modulation of neuronal activity without interfering with our ability to record neural activity in chronically implanted animals (Sunderam, et al, 2006), and therefore can apply feedback stimulation. With head acceleration measurements, we are able to determine state of vigilance (Sunderam, et al, 2007). The aims of this grant are three fold. First, to investigate if the addition of state of vigilance as a discrimination feature improves identification of preseizure states. Second, to implement and test an active probe of brain state through small amplitude stimulations to detect changes in brain state indicative of a preseizure state. We expect after implementing both the passive and active prediction methods that we will still observe significant false prediction rates. The third aim is to probe through stimulation the nature of these detections (a) if the false predictions are simply misclassifications OR (b) if the identified preseizure state is seizure permissive - a state that only sometimes transitions to seizure - and the 'false predictions' are correct identifications of this state. From a basic science standpoint, this should give insight into the seizure generation process and long-term treatment. From a more practical short-term application standpoint, detection of a seizure permissive state and the relevant transition probabilities will have great utility in the development of a useful feedback intervention. Specifically, one then targets intervention - for example electrical stimulation - in response to detection of the state to modify this transition probability. The extension of this work in future years will be to test a range of responsive stimuli to preseizure detections for their ability to prevent seizure.